Research map / Label-efficient and robust learning / Recognition and long-tail
DriveMamba: Task-Centric Scalable State Space Model for Efficient End-to-End Autonomous Driving
Haisheng Su, Wei Wu, Feixiang Song, Junjie Zhang, Zhenjie Yang, Junchi Yan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add view correspondence learning layer (ablation) | DriveMamba-Tiny · nuScenes val planning | avg L2 (m): 0.61 -> 0.44 | Improved |
| add long-term temporal fusion (ablation) | DriveMamba-Tiny · nuScenes val planning | avg collision (%): 0.18 -> 0.15 | Improved |
| add task relation modeling layer (ablation) | DriveMamba-Tiny · nuScenes val planning | avg collision (%): 0.20 -> 0.15 | Improved |
| joint modeling with single B-Mamba layer instead of divided modeling (ablation) | DriveMamba-Tiny · nuScenes val planning | avg L2 (m): 0.46 -> 0.44 | Improved |
| add iterative optimization (ablation) | DriveMamba-Tiny · nuScenes val planning | avg collision (%): 0.38 -> 0.15 | Improved |
| add residual learning (ablation) | DriveMamba-Tiny · nuScenes val planning | avg collision (%): 0.17 -> 0.15 | Improved |
| hybrid scan (H/V alternation + trajectory-centric) vs single scan types (ablation) | DriveMamba-Tiny · nuScenes val | detection NDS / planning L2: NDS 45.4 vs 35.1 (TC only); L2 0.44 vs 0.55 (H-first) | Improved |
| trajectory-centric instead of ego-centric local-to-global scan (ablation) | DriveMamba-Tiny · Bench2Drive closed-loop | Driving Score: 50.27 -> 53.54 | Improved |
| end-to-end training instead of two-stage training (ablation) | DriveMamba-Tiny · nuScenes val planning | avg collision (%): 0.23 -> 0.15 | Improved |
| scale decoder layers 3 to 12 (ablation) | DriveMamba (ResNet-50) · Bench2Drive Dev10 / nuScenes perception | Driving Score / mapping mAP: DS 51.1 -> 66.5; mapping mAP 50.3 -> 46.6 | Mixed |
| scale visual encoder from ResNet-50 to ViT-L/16 (3 decoder layers) (ablation) | DriveMamba · nuScenes detection / Dev10 planning | detection mAP / DS: mAP 34.8 -> 50.8; DS 51.1 -> 55.8 | Mixed |
| vary initialization of trajectory prior (Origin/Uniform/Random) (ablation) | DriveMamba-Tiny · nuScenes val | planning L2 / collision: L2 0.44 / 0.46 / 0.45 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DriveTransformer-Large — DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving (ICLR 2025) · better on 3 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving (ICLR 2025 · similarity 0.91)
- DualAD: Disentangling the Dynamic and Static World for End-to-End Driving (CVPR 2024 · similarity 0.88)
- Trajectory Mamba: Efficient Attention-Mamba Forecasting Model Based on Selective SSM (CVPR 2025 · similarity 0.88)
- Navigation-Guided Sparse Scene Representation for End-to-End Autonomous Driving (ICLR 2025 · similarity 0.87)
- VR-Drive: Viewpoint-Robust End-to-End Driving with Feed-Forward 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.86)
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
